Hartford HealthCare has introduced PatientGPT, an AI-powered chatbot embedded in its patient portal that interprets lab results and answers health questions using patients' own medical records. The tool, built with clinical AI company K Health, arrives as roughly 32% of U.S. adults report turning to AI for health advice.
PatientGPT launched as a beta in March and has logged 8,300 conversations from 6,000 unique patients. Patients must manually grant permission for the tool to access their data. "It's HIPAA-compliant, it's secure, it's private, and it's yours," said Dr. Ajay Kumar, chief clinical officer at Hartford HealthCare.
The deployment mirrors a broader push to weave AI for Healthcare directly into patient-facing experiences, not just back-office functions.
How PatientGPT personalizes responses
Unlike general-purpose AI chatbots, PatientGPT pulls from a patient's actual medical history-lab results, appointment records, and clinical interactions-to generate tailored answers. "This particular system knows you," Kumar said. "It knows how many times you've interacted with the doctors and Hartford HealthCare. It knows the future appointments you might have with a different type of specialist or the blood test… the doctor asked you to do off time." The tool can explain lab results in plain language, flag potential treatment interactions, summarize clinical conversations, and help book virtual or in-person appointments.
Built-in safeguards and escalation pathways
The chatbot cannot prescribe medications or recommend specific treatments. When a question exceeds its scope, built-in "escalation pathways" steer patients toward scheduling a clinical encounter. "You can imagine what an empowering tool it would be for patients to be able to get a lot of questions answered, and when the things become bit more less clear, to allow the physicians to interact," Kumar said. All interactions sit inside the patient portal, under the same login controls that protect the rest of a patient's medical record.
Early adoption and rollout plans
A Kaiser Family Foundation survey found that 32% of adults have used AI for health information, signaling strong demand for patient-facing tools like this one. At Hartford HealthCare, early conversations most often include follow-up questions, general health information, and medication eligibility checks-such as for GLP-1 drugs. Kumar said patients arriving for appointments now appear "well-informed," leading to "more impactful conversations" about next steps. The health system plans to extend PatientGPT to over a million patients in the next four to six weeks, with multilingual versions in development.
Why this matters for healthcare professionals
When patients come to appointments with questions already answered, clinicians can spend less time on routine explanations and more on complex decision-making. A tool that summarizes patient inquiries and flags concerns in advance could also reduce administrative strain. As more health systems test patient-facing AI, clinicians will likely encounter a well-read patient population that expects clear, efficient follow-through-making digital triage and message discipline a growing part of clinical workflow design.
Your membership also unlocks:
